Abstract
The lack of systematically collected and analysed data about the effect of telemedicine on patient-provider communication is a frequently cited barrier for why video communication has yet to reach its full potential. Existing research provides little information about the subtle and detailed changes in communication that take place over video. Comprehensive investigations of actual medical encounter behaviour are therefore required, including verbal content analysis, which uses interaction analysis systems (IAS) to describe and categorize the communication that has taken place. Ten IAS studies were identified in the literature. Although it is difficult to generalize due to differences in methodology and context, some tentative conclusions can be drawn. First, on-site providers tend to be substantially less active than off-site providers, suggesting that the former typically serve as facilitators and observers, rather than active participants. Second, just as in the conventional face-to-face setting, providers' utterances tend to predominate in telemedicine. Third, conventional patterns of more task-focused than socio-emotional utterances tend to persist in telemedicine. However, some studies found telemedicine to be more patient-centred than conventional medicine, and others found it less so. We do not yet have a full understanding of the subtractive and enhancing effects of telemedicine on provider-patient relations and outcomes.
Introduction
Despite growing use of telemedicine internationally, 1 there is a general feeling that it has not yet reached its full potential. A large proportion of rural and urban communities that could benefit continue to lack access to telemedicine; so too do many developing nations. Furthermore, telemedicine programmes tend to be short-lived while those that do survive frequently experience disappointing levels of usage. A number of factors contribute to this state of affairs, including financial barriers, ethical and legal concerns, piecemeal development of the telecommunications infrastructure and lack of resources to sustain use, particularly in some regions and among certain vulnerable populations. 2–18 A dearth of systematically collected and analysed evaluation data regarding the effects of telemedicine on cost, quality and access is another often cited barrier. 19–25 This is also true of telemedicine's effect on provider-patient communication, which, according to a number of observers, should be an important component of any investigation in this area. 26–30
Understanding the effect of telemedicine on the provider-patient relationship is especially important in view of the marked changes in the technical and interpersonal contexts within which communication takes place. 31 Such knowledge would be especially helpful in overcoming resistance to telemedicine, thereby promoting further use where appropriate. Varying degrees of uncertainty exist regarding how providers and patients should behave during teleconsultations, leading to a certain degree of hesitancy, anxiety and conflict as the participants attempt to become familiar with the technique. There is a long history of developing relatively simple and inexpensive interventions to influence provider-patient behaviour in conventional medical settings. 32,33 Similar interventions would better prepare patients and providers for telemedicine.
Telemedicine could be adapted to existing behavioural patterns through improvements in the human/technology interface. 34 The more accommodating the interface is to prevailing consultation norms the more acceptable telemedicine is likely to be to patients and providers. Not only would more user-friendly equipment reduce patient and provider concerns about an unfamiliar consultation modality, but it would also facilitate incorporation of telemedicine into existing medical practice more generally. 27,35 Knowledge of telemedicine's impact would also increase user acceptance.
The aim of the present review is to summarize prevailing knowledge about provider-patient communication during telemedicine encounters.
Limitations of the existing knowledge base
Comparatively little research has been conducted to investigate the effect of videoconferencing on provider-patient relations. There have been several literature reviews, however. Whitten and colleagues systematically reviewed studies examining satisfaction with telemedicine, most of which relied on simple survey questionnaires administered after teleconsultations had taken place. 36,37 Despite methodological and conceptual weaknesses with most of the studies examined, Whitten and colleagues concluded that the research reveals high levels of patient satisfaction with telemedicine, particularly with respect to travel, waiting time and access to comprehensive specialist care, but ‘some disquiet’ in the area of provider-patient communication. This is in contrast to providers who, while also generally positive, appear to hold more reservations, particularly with respect to telemedicine's capabilities and uses. It was posited that patients may appreciate telemedicine more because they benefit most immediately in terms of travelling and scheduling. Providers, on the other hand, may need to see greater benefits; otherwise, the additional training, technical requirements and implementation costs associated with using telemedicine may be viewed as an unnecessary intrusion into their practices.
Miller 38 focused more explicitly on the doctor-patient relationship, reviewing 38 studies, most of which were, again, post-encounter surveys of medical encounter participants, that examined provider-patient communication under telemedicine in a variety of specialty areas, including psychiatry, otolaryngology and dermatology. The findings from each study were coded according to 23 categories and a positive and negative rating assigned to each of 213 communication results. Approximately 80% of abstracted findings favoured doctor-patient interaction, with all but two of the 23 categories analysed (non-verbal behaviour and lack of touch) reporting more positive than negative results. Attributes that were ranked especially favourably included both patient and provider comfort, understanding and explanation, patient-provider relations, communicative efficacy, rapport development, embarrassment, anxiety/nervousness, audio quality, video quality, patient involvement and multiple providers. This more general review was subsequently repeated with 57 studies that examined the nature and content of provider-patient communication in the area of telepsychiatry where more has been written than, perhaps, any other telemedicine specialty area. 39 Here, findings from each study were coded according to 23 categories and a positive and negative rating assigned to each of the 550 communication results abstracted. More than 75% of the findings favoured telemedicine. Non-verbal behaviour was the only area for which the number of positive findings did not exceed the number of negative findings.
Results from recent post-encounter studies are consistent with the broader literature documenting general satisfaction with telemedicine. 40–42 The studies show high levels of satisfaction and acceptance with telemedicine, both when evaluated independently and in comparison to conventional in-person encounters. In fact, there is evidence to suggest that patients may be especially satisfied with the convenience that telemedicine offers, both with respect to travel and appointments and with respect to being able to consult with particular types of providers, say, for example, those who share their own culture and language. Not only might remote consultation be less stressful for patients who otherwise need to travel long distances for their appointments but it may allow for better patient and provider matching than otherwise possible.
Although the evidence appears to favour communication via telemedicine, the generalizability of these results may be limited. This is because most studies included in the aforementioned reviews focus more on overall system performance and satisfaction with telemedicine attributes rather than communication. Furthermore, most rely on post-encounter surveys of medical encounter participants where patients typically report high levels of satisfaction with the care received. 43 Post-encounter surveys also provide little information about the more subtle and detailed changes in communication that take place via videoconferences and how such changes may affect patient outcomes over time.
Towards an improved understanding of the provider-patient dynamic
To obtain a fuller understanding of the effects of telemedicine on the provider-patient relationship, including insights unavailable to researchers relying solely on retrospective participant assessments, Miller 38 suggested that post-encounter surveys of medical encounter participants could be supplemented with more comprehensive examinations of actual medical encounter content. For example, verbal content studies have long been used to study the way providers and patients communicate in conventional medical settings and how those behaviours correlate with patient, provider and system-level attributes and outcomes. 28,44–47 Interaction analysis systems describe and categorize communication behaviours. Most such systems employ an exhaustive taxonomy for classifying verbal utterances or ‘the smallest meaningful and distinguishable speech segment, conveying only one thought or relating to one item of interest.’ 47
The most widely used verbal content analysis instrument is the Roter Interaction Analysis System (RIAS), which employs 34 categories to describe physician behaviour and 28 categories to describe patient behaviour. 48 The RIAS codes both socio-emotional or affective (care-oriented) behaviour and instrumental or task-focused (cure-oriented) behaviour. Examples of socio-emotional exchange include showing concern, social behaviour and disagreement. Examples of task-focused exchange include giving information, asking questions and counselling. Although the RIAS primarily focuses on verbal behaviour, voice tone and intonation are accounted for as well. The RIAS also asks coders to rate the general emotional context of each encounter. Coding may be accomplished using videotape, audiotape or literal transcripts. Like other interaction analysis instruments, the RIAS has been shown to be reliable. 49 Recent work has adapted the RIAS to telemedicine, specifically. 50,51
Interaction analysis studies
Peer-reviewed articles were identified by searching all years included in the PubMed database. Combinations of the following keywords were used: communication, physician-patient relations, provider-patient relations, televideo and telemedicine. An article was included in the review if it reported the results of an interaction analysis study. Approximately 300 articles were found, of which 10 used interaction analysis to study provider-patient communication during teleconsultations. 51–60 Further details are provided in Table 1.
Telemedicine and provider-patient communication: interaction analysis studies
Nelson et al. 51 used an adapted version of the RIAS to examine the verbal content of 46 teleconsultations that took place between an on-site patient and nurse/social worker and one of a number of different types of off-site specialists based at Kansas University. Conventional RIAS categories proved reliable in telemedicine, as did an aggregate technology-related category, although there were few technology-related utterances overall (<1% of total). Most utterances were made by providers (57%), followed by patients (27%), family members (10%) and presenters (6%). There were considerably more task-focused utterances than socio-emotional utterances. Patients most commonly gave information (62%), showed agreement (13%) and made personal remarks (6%). Providers most commonly asked closed-ended questions (18%), showed agreement (17%) and gave information (14%). Nelson et al. concluded that limited technology-related utterances imply a certain degree of comfort with videoconferencing.
Agha et al. 52 used the RIAS to compare the verbal content of eight in-person consultations and 11 teleconsultations between an on-site patient and nurse and an off-site pulmonary specialist based at the Milwaukee Veteran Affairs Medical Center. On average, there were equal proportions of utterances by physicians and patients with in-person visits (46% each). However, physicians accounted for more utterances than patients under telemedicine (48% vs. 38%). Nurses contributed 6% of total utterances in telemedicine; companions 7% in telemedicine, 9% in person. There was far more biomedical than psychosocial information exchanged during both types of consultations. Physicians were more likely to use orientation statements during in-person visits. Patients were more likely to make requests for repetition during telemedicine. Agha et al. concluded that telemedicine visits are more physician-centred than in-person consultations, with physicians controlling the dialogue and patients assuming a more passive role.
Wakefield et al. 60 used the RIAS to compare the verbal content of 42 telephone consultations with 42 video consultations between patients at home who had previously been hospitalized for heart failure and an off-site nurse case manager based in Iowa. On average, nurses made slightly more utterances than patients during both video- and telephone consultations (52% vs. 48%). In general, nurses were more likely to gather data, build relationships and partnerships, whereas patients were more likely to give information. Nurses, however, were more likely to ask open-ended questions, make back-channel responses (indicating listening), friendly jokes and check for understanding with telephone visits; compliments given and partnership were more common with video. Patients were more likely to give lifestyle information and approval comments with telephone visits; closed-ended questions were more common with video. Wakefield et al. concluded that the value added by using low-cost videophone technology did not appear to be worth the additional complexities.
Liu et al. 56 used verbal content analysis to compare 20 in-person consultations to 20 teleconsultations in Gunma University Hospital in Japan. The results showed significantly more utterances with in-person than telemedicine consultations, with patients making more utterances than physicians in both settings (56% vs. 44%). They also showed significantly more conversational turns during in-person visits and greater requests for repetition during telemedicine. Physicians were less likely to make facilitation, empathy or praise utterances during telemedicine visits. There were no differences in closed- or open-questions asked. Liu et al. concluded by suggesting a new training programme to improve doctors' communication skills and their ability to express empathy during telemedicine.
Demiris et al. 55 used verbal content analysis to compare 40 in-person consultations to 54 teleconsultations between an on-site patient and an off-site dermatology clinic in Missouri. Smalltalk took place in 20% of in-person consultations and 30% of telemedicine visits. Clinical assessment took place in all consultations. Although the difference was not significant, 90% of in-person visits included a patient education component compared with 78% of telemedicine visits. Similar percentages of telemedicine and in-person visits addressed treatment, compliance, psychosocial and administrative matters. Technical matters were raised in 15% of telemedicine visits. Demiris et al. concluded that communication patterns in telemedicine and in patient visits were similar.
Demiris et al. 54 used patient and provider self-reports to assess 122 virtual visits between chronically ill elderly patients at home and off-site nurses based at one of three Minnesota home care agencies. They also used verbal content analysis to examine a subset of 30 of these visits. Technical quality was given an average rating of 95 out of 100; problems in establishing a connection were recorded on 8% of occasions. The highest proportion of time was spent assessing patients' clinical status (42%), followed by compliance (13%), psychosocial matters (10%), and education and informal talk (both 8%). On average, nurses spoke for 59% of the time and made 67% of the utterances. Demiris et al. concluded that technical problems do not interfere with the care provided during virtual visits.
Tachakra and Rajani 59 used verbal content analysis to compare 30 in-person consultations with 30 teleconsultations between an on-site patient and nurse practitioner and an off-site physician at a minor accident and treatment service in the UK. They found that in both doctor-patient and doctor-nurse communication there were more words and higher rates of turn taking, interruptions and back-channel responses with telemedicine than in-person consultation. There was little difference in patient-nurse communication between the two settings. Tachakra and Rajani concluded that telemedicine empowered patients to ask more questions while the doctor took greater care to achieve coordination of beliefs with patients due, perhaps, in part to the lack of multisensory feedback.
Savenstedt et al. 57 used verbal content analysis to examine 15 teleconsultations between an on-site nurse based at a geriatric nursing home and an off-site geriatrician based at a university hospital in northern Sweden. They found that most problems or tasks (69%) could be dealt with either by telephone or teleconsultation. They also indicated that the teleconsultations approximated that of a conventional ward round. The behaviours analysed included: (1) the nurses' presentation of the problem and tasks (how they were presented, types of questions asked, aim); (2) the physician's response (how, types of questions); and (3) use of videoconferencing (10% of consultations did not fully exploit it). Savenstedt et al. concluded that by requiring more systematic presentation and more preparation, teleconsultation improves the structure of work but that mutual trust is important to the success of the interaction.
Street et al. 58 used verbal content analysis to examine 26 teleconsultations between an on-site patient and primary care practitioner (PCP) and an off-site physician specializing in one of a number of different areas based at the Texas Tech Health Science Center. They found very little group discussion among participants, with most talk occurring between the doctor and PCP or patient and little talk between the PCP and patient. Most utterances were made by the specialists (45%), followed by the PCPs (34%) and patients (23%). Information giving was more equally distributed, although patients received the least (17%) and specialists the most (55%). Street et al. concluded that, overall, the specialists dominated, asking the most questions, exerting the most control and being talked to the most often. In contrast, patients were the least active, making the fewest utterances, asking the fewest questions, exerting the least control and receiving the least amount of information.
Ball et al. 53 used verbal content analysis to compare visual (in person, video) and non-visual (telephone, hands-free telephone) psychiatric examinations received by each of six patients in London. The results suggested that patients were more anxious in non-visual modes where they tended to adopt the least relaxed body postures. Irrespective of the mode, the doctors' angle of recline was always greater than that of the patients, while participants displayed high mutual gaze both in person and over video. No differences in verbal content measures indicative of partnership building, information giving, or question asking could be discerned. Ball et al. concluded that visual cues are important to both patients and doctors, but that in-person consultations are not the only way to provide them.
Implications
There are substantial methodological differences in the studies reviewed. The findings from several studies are consistent with provider-patient research in conventional, face-to-face settings, while others are not.
Heterogeneity in study setting
There is substantial heterogeneity in the ten interaction analysis studies, both in regard to study setting and design. Each study focused on a different specialty area, including psychiatry, 53 dermatology, 55 emergency medicine, 59 pulmonary care, 52 internal medicine, 56 gerontology 57 and home care. 54,60 There were also two that included consultations in multiple specialties. 51,58 Whereas some studies included a second, on-site provider of various types (i.e. a GP, nurse, nurse practitioner or physician assistant) (four studies) 51,52,58,59 and occasionally even a family member or other companion (two studies), 51,52 others did not (six studies). 53–57,60 Furthermore, the consulting provider was typically a physician (eight studies) 51–53,55–59 but occasionally a nurse (two studies). 54,60 To varying degrees, studies focused explicitly on analysing partial/entire visits, initial/follow-up encounters, and/or participants with no/some previous experience with the technologies used. This variation makes it difficult to develop conclusions regarding the effects of telemedicine on provider-patient communication within a given category of interaction defined by specialty, number and types of participants, and other consultation characteristics. On the other hand, the variation provides opportunities to identify themes that cut across different settings, thereby permitting us to draw lessons applicable to different teleconsultation contexts.
Heterogeneity in study quality
Heterogeneity in study quality is reflected in several areas. Few used the RIAS (three studies) 51,52,60 most another, typically unspecified instrument (six studies). 53–59 Moreover, the number of categories abstracted differed, ranging from 30 or more categories in some investigations (three studies) 51,52,60 to 13-17 categories 53–57 to less than 5 (two studies). 58,59 Only one study 53 assessed non-verbal content even though most research has identified it as potentially the most difficult aspect of provider-patient communication during telemedicine. 38,39 Some effort was made to ensure data quality/reliability in most studies, 50,51,54,55,57,58,60 though not in all. 53,56,59 In contrast most studies were descriptive only; 50,53,54,56–58,60 few sought to evaluate specific hypotheses. 51,55,59 Some studies employed one or more comparison groups, comparing telemedicine to in-person visits 52,53,55,56,59 or telephone visits 53,60 while others did not include a comparison group in their design. 51,54,57,58 Inclusion of a comparison group is essential for negating single group threats to internal validity. The best way to address the confounding effects of selection in multiple group designs is to assign participants randomly to one consultation situation (telemedicine) or another (in person). No study did so, although two used the same participants for both. 53,56 Overall, the number of consultations analysed were relatively small (range 6–54). This suggests that the studies reviewed may not have had sufficient power to detect important differences in communication across the consultation modalities analysed. Thus, conclusions regarding similarities between telemedicine and in-person consultations may have been different had larger sample sizes been used.
Variation in encounter length
Some studies found that consultations took longer via telemedicine 59 while others found that consultations were longer in person. 55,56 Some studies found that there was no difference between the two modalities. 52 Variation in length may depend on the technical aspects of the consultations in question. For example, the bandwidth, frequency of technical problems and degree of familiarity with the technology used may all be relevant factors. Indeed, the greater the novelty the more likely that participants will spend time negotiating their roles and the context within which the consultation takes place. 61 It is also likely that such negotiation would be even more difficult during the triadic medical encounters typical of telemedicine, particularly since they are considerably less common than the conventional doctor-patient dyad with which most people have become accustomed face-to-face. Variation in consultation length may also derive from the interpersonal aspects of the consultation situation; for example, whether the consultation is an initial or follow-up visit and whether participants had met previously or not. Indeed, evidence suggests that how communication unfolds within teleconsultations depends, in part, on the level of familiarity that exists among the participants involved. 62,63 Those that have met previously are likely to communicate quite differently, say, with more social conversation (and perhaps, greater warmth) than those who had never met before.
Role of on-site providers
On-site providers were substantially less active than off-site providers, 51,58,59 a finding reinforced by other investigations. 62,64 This suggests that presenters typically serve as facilitators and observers, rather than active participants. Post-encounter surveys reveal positive feelings regarding the role of multiple providers during teleconsultations, 36–38 but the dynamic is more complicated than suggested by these studies. The presence of a second provider might increase patient confidence, say, because ‘two doctors are better than one.’ 31,64 However, passive engagement could prove frustrating for consulting providers who, under certain circumstances, might prefer more active involvement on the part of their on-site colleagues, which, in addition to facilitating the interaction and advocating for the patient might include making up for the loss of non-verbal cues and mediating consultants' therapeutic influence. 31,62,63 An on-site provider could also complement the strengths of off-site consultants. Specialists tend to focus primarily on biomedical issues. In contrast, generalists, whether nurses and other primary care practitioners, tend to adopt a broader view of health that also includes a concern for patients' psychosocial environments. 62
Consultant dominance of visit dialogue
Just as in the conventional face-to-face setting, provider utterances tended to predominate in telemedicine encounters, with fewer utterances being made by patients, on-site providers and family members, 51,52,58 although, in a single study relying on simulated encounters, patient utterances tended to predominate. 56 There is often pressure to use telemedicine systems as productively and efficiently as possible. This may lead some clinicians to provide patients with fewer opportunities to participate. Less active patient involvement is likely to lead to greater emphasis on clinical information exchange at the expense of psychosocial exchange, possibly compromising patients' psychological state and satisfaction, not to mention the ability of providers to detect social problems. Patient-centred communication may be at risk.
Task-focused vs. socio-emotional exchange
The conventional pattern of more task-focused than socio-emotional utterances tends to persist in telemedicine. Roter and Hall 32 concluded that during the typical in-person encounter, physicians spend most of their time giving information (35%) and seeking information (23%) followed by positive talk (15%), partnership building (11%), social conversation (6%) and negative talk (1%). Patients, in contrast, typically spend most of their time giving information (47%) and comparatively little asking questions (7%). Patients also tend to spend more time talking both negatively (8%) and positively (19%) as well as socially (13%). Although the distribution of behaviours varied, biomedical, task-focused exchange tended to predominate as with conventional face-to-face settings. This is reflected both in the telemedicine-only studies reviewed 51,58 as well as in those studies showing no significant differences in behaviour between telemedicine and in-person encounters. 53,55 Other comparisons, however, indicated that telemedicine might be less patient-centred than conventional contact, 52 and some that it might be more so. 59
Future directions
Several observers have called for more scientifically robust research examining the relationship between telemedicine and provider-patient communication. 21,28,29,36,38 These calls typically include the application of interaction analysis and other instruments to electronically mediated communications to better understand the ways in which providers and patients communicate both verbally and non-verbally. However, despite the potential importance of interaction analysis for improving telemedicine practice and outcomes, only a handful of studies have been conducted to date. There are two main reasons. Research has focused primarily on assessing the relationship between telemedicine system attributes and cost, quality and access. This is understandable, particularly during the early evolution of the field. However, this emphasis comes at the expense of more in-depth investigations examining the mechanisms underlying the relationships identified, including the way providers and patients communicate with one another. This suggests a need to focus more on explaining why particular patient, provider and contextual characteristics tend to be associated with certain outcomes. Doing so, however, is only a necessary condition for stimulating more research in this area. This is because interaction analysis studies are very resource intensive, requiring considerable data collection and analytical capabilities to complete. Thus, more interaction analysis research will require substantial additional funding from both governmental and non-governmental sources.
Further interaction analysis research is particularly important because those studies identifying similar relationships between telemedicine and conventional care would inform arguments supporting greater user confidence, systems development, reimbursement and liability protection, among other things. Further research is also important because studies identifying differences between telemedicine and conventional health system contact would enable researchers to address potential difficulties in provider-patient interaction before widespread implementation of particular telemedicine technologies took place.
In addition to comparing telemedicine to in-person interactions, it is important that further research compares the content and nature of provider-patient communication during teleconsultations across different specialties and technical specifications. 50 Telemedicine encompasses several different forms of information transmission, communication technologies and user interface. Since use of these technologies has evolved rapidly, particularly with respect to bandwidth, image quality, peripheral devices and user interfaces, it is important that researchers investigate the relationship between telemedicine and provider-patient communication among the most common technological configurations. This information would be particularly useful for providers, whether in establishing new telemedicine clinics or instituting electronically mediated interventions in conventional face-to-face settings. Knowing how different technologies will affect the provider-patient relationship will also better enable system developers to select the most clinically effective, cost-efficient technologies possible. It would enable researchers to tailor patient and provider interventions to the specific technologies used as well.
Once patterns of communication are understood, researchers should examine the relationship between these patterns and patient, provider and contextual characteristics, as well as important outcomes such as satisfaction, adherence and compliance, health and clinical status, recall and understanding, and psychological wellbeing. 28 Knowing what interaction patterns lead to the most desirable outcomes would inform the development and implementation of more effective behavioural interventions and user interfaces. In addition, more rigorous research designs need to be employed. This includes randomization of study participants to a telemedicine and an in-person control group if possible, or selection of a suitable comparison group if not. It also includes the collection of comprehensive baseline data to better account for the potentially confounding effects of participant differences across study groups. Well-known and validated interaction analysis instruments such as the RIAS should be used, not to mention similar instruments that describe non-verbal behaviour. Careful attention should be paid to sample size to ensure that sufficient power is available to detect differences, not only across consultation setting but also across relevant subsets of patients and providers. Ultimately, high quality research is necessary if we are to fully understand the subtractive and enhancing effects of telemedicine on the provider-patient relationship.